Confirmations and payment reminders.
Check-in instructions, Wi-Fi information, checkout reminders.
Cleaning notifications. Maintenance coordination. Review requests.
It looks like the perfect place to use AI.
Most of it isn't.
Predictable problems
don't need intelligence
A reservation already gives us a remarkably useful structure:
Booked
The stay exists.
Pre-arrival
Instructions go out on a schedule.
Check-in
Access information when they need it.
Stay
The guest is in the house.
Checkout
We know when it is supposed to happen.
Post-stay
Ask for the review.
We know when the guest is arriving.
We know when checkout is.
We know the property they booked.
And we know much of the information they'll need at each stage.
So Hospitable handles much of the guest journey with ordinary marketing automation.
Send the appropriate instructions.
Send access information.
Send checkout instructions.
Request a review.
There is no reason to ask an LLM what to do.
If the trigger is known, the context is known, and the action is known, use a rule.
The same applies
behind the scenes
The guest isn't the only person who needs information.
need to know when a property is ready for turnover.
need to know what needs attention.
Those people don't need to log into an AI platform — or any platform at all.
They use text messages.
So that's where the system meets them.
n8n acts as the deterministic orchestration layer, moving information between the reservation system, operating workflows and the people responsible for the property.
Known checkout tomorrow? Send the appropriate reminder.
Confirmed turnover? Text the cleaning crew.
Known maintenance task? Send the right information to the right person.
The workflow may cross several systems, but the logic is still predictable.
AI doesn't improve a decision we've already defined.
AI belongs where
the structure disappears
Then the guest starts talking.
“We had an amazing weekend. We're heading out now. Thanks for everything!”
Humans immediately understand what that means operationally.
Software doesn't.
The guest hasn't changed a status field or clicked a checkout button. They've expressed an operational event in natural language.
That's where AI becomes useful.
An LLM can interpret the conversation and transform it into structured information:
{
"guest_status": "DEPARTED",
"confidence": "high",
"next_action": "BEGIN_TURNOVER"
}
Now n8n has something deterministic to work with again.
Conversation
The guest talks in their own words.
AI
Interprets the message.
Event
A structured operational state.
Workflow
Deterministic orchestration.
Text
The people who need it get it.
AI bridges the gap between the messy way humans communicate and the structured way operational systems work.
Then it gets out of the way.
The architecture is
intentionally boring
Most of the system is conventional automation.
That's a feature.
manages predictable guest communications.
orchestrates deterministic workflows across systems.
delivers operational information to cleaning and maintenance crews where they already work.
interprets the unstructured moments that can't be reliably expressed as rules.
handle decisions where judgment still matters.
The goal was never to maximize the amount of AI in the system.
It was to put intelligence exactly where the deterministic system needed it.
That's the idea
behind Big Splash
It's easy to start an AI project by asking:
What can we automate with AI?
Which parts of this problem are actually uncertain?
Trigger and action knownAutomate it.
Language needs structureUse AI.
If judgment matters: keep a human involved.
AI is incredibly powerful.
Knowing where not to use it is part of building an intelligent system.